Top ShopA digital installation democratizing fashion
Fashion cannot be elite if it is to come into fashion. For Topshop to similarly include their clients in what comes next, access is as important as the showcase designs themselves. In this project, social media was leveraged through video to achieve exactly that. For London Fashion Week, the Topshop catwalk was enhanced with a stream of impressions from the event itself. Individual video frames, making up the stream of the show shown on public screens, were built from a mosaic of Facebook and Instagram pictures marked for use with the Topshop hashtag.
A crowd of fashionitas invited to post their own enthusiastically made pictures from within the show venue, collaboratively created the real-time mosaic video. A three-display set-up, installed at the brand’s flagship store on Oxford Street in London to share the experience with the public at large, had a center screen showcasing the interactive mosaic video itself. Viewers could interact with the images from the street via a motion-based interface to explore the individual elements. The two side screens showcased the virtually never-ending selection of source imagery.
Selected fashion bloggers curated the stream of incoming photos and looked for quality and fair representation of the seasonal silhouettes. We created the application that allowed these human actors to quickly move images into algorithms. To fit the pictures into the mosaic version of the live video we had to calculate both source and destination characteristics for the individual tiles making up the image, which requires speed and performance. We showed of both in spades.
To collect the photos themselves, we developed a node.js-based application crawling Instagram for tagged images. The meta data was stored in a MongoDB database for the curator application to tap into. Traditional mosaic images primarily focus on calculating a descriptor from the source images. We had to equally focus on describing the destination grid in real time, if we were to keep up with live video. High performance custom C++ framework code and GPU algorithms ran simultaneous threads to match the selected posts with the incoming frames, and even more importantly, the fashion of the season.
to the team
- PETER HELLICAR & JOEL LEWIS
- AMALIE ENGLESON